The first layer of the visual cortex (and what the input layers convolutional kernels in visual NN converge to) are those Gabor kernels - cosine multiplied by

The first layer of the visual cortex (and what the input layers convolutional kernels in visual NN converge to) are those Gabor kernels - cosine multiplied by exponentially decreasing amplitude thus de-facto limiting the spatial attention of the given neuron to a spot.